Database / Milvus Vector database Interview questions
How do you configure replicas in Milvus for read scalability?
Replicas are configured at load time, specifying how many independent copies of a collection's (or specific partitions') segments should be loaded across separate sets of Query Nodes.
client.load_collection( collection_name="products", replica_number=3 )
QueryCoord handles distributing each replica's segments across available Query Nodes and load-balancing incoming search requests across the replicas, so from the client's perspective, increasing replica_number mostly just means the collection can handle more concurrent search traffic and tolerate the loss of some Query Nodes without becoming unavailable for search.
The practical constraint is resource availability: each additional replica requires enough Query Node memory and compute capacity to hold another full copy of the loaded segments, so scaling replica count effectively requires scaling the underlying Query Node fleet's total capacity alongside it, rather than replica count being a free lever independent of cluster resourcing.
More Related questions...
